Hedamycin and rubiflavin complexes with deoxyribonucleic acid and

Studies Toward the Total Synthesis of Pluraflavin A. John Hartung , Benjamin J. D. Wright , Samuel J. Danishefsky. Chemistry - A European Journal 2014...
1 downloads 0 Views 1MB Size
BIOCHEMISTRY

Hedamycin and Rubiflavin Complexes with Deoxyribonucleic Acid and Other Polynucleotides* Helen L. White? and James R. White ABSTRACT: The related antibiotics rubiflavin and hedamycin interact similarly with deoxyribonucleic acid in citro. Formation of strong deoxyribonucleic acid-antibiotic complexes brings about a number of changes in the physical and chemical properties of the deoxyribonucleic acid. As a result of complex formation the visible absorption spectra of the antibiotics are red shifted and depressed, with development of shoulders suggesting that the antibiotic chromophores are in a hydrophobic environment. The temperature of the deoxyribonucleic acid melting transition is increased in the presence of the antibiotics, and enzymatic digestion of native deoxyribonucleic acid by snake venom phosphodiesterase is strongly inhibited. Exhaustive dialysis of Escherichia coli deoxyribonucleic acid-rubiflavin complexes showed that at low ionic strength the ratio of tightly bound rubiflavin molecules per deoxyribonucleic acid nucleotide is 0.53, while for hedamycin the ratio is 0.20. Denaturation of deoxyribonucleic acid did not greatly affect the tight binding of rubiflavin (ratio 0.57), but high ionic strength reduced it (ratio 0.23). Raising the percentage guanosine plus cytosine by using Micrococcus lysodeikticus deoxyribonucleic acid had little effect on the amount of tight binding (ratio 0.57), while for poly dAT, binding was decreased significantly (ratio 0.3.5). Complexes of D N A with rubiflavin and hedamycin band in CsCl gradients at positions of considerably lower density than the uncomplexed DNA. Mixtures of

R

ubiflavin, C ~ ~ H ~ Q - ~ isI NanOantitumor ~, and bactericidal antibiotic isolated from Streptomyces (SC 3728) (Aszalos et al., 1965). We have previously reported (White and White, 1967) that rubiflavin causes a specific inhibition of D N A synthesis in Escherichia coli and that a portion of the D N A isolated from a rubiflavin-treated culture by isopycnic centrifugation bands at a lower density than native DNA. Various ~

* From the Department of Biochemistry, School of Medicine,

1030

University of North Carolina, Chapel Hill, North Carolina 27514. Received August 9, 1968, This work was supported by a research grant from the U. S. Public Health Service (AI-06211) and an equipment grant from the National Science Foundation (GB-4577). t Predoctoral Fellow of the National Aeronautics and Space Administration. This work was done in partial fulfillment of the requirements for the degree of Doctor of Philosophy in Biochemistry, from the Department of Biochemistry, School of Medicine, University of North Carolina at Chapel Hill. Present address: School of Pharmacy, University of North Carolina, Chapel Hill, N. C. 27514.

WHITE AND WHITE

the complexes with the uncomplexed D N A in CsCl solution form separate bands upon centrifugation, indicating that equilibration of antibiotic molecules among individual DNA molecules is not extensive. The binding is probably noncovalent, since rubiflavin can be extracted with benzene from an aqueous solution of the complex at high ionic strength. At least three kinds of binding to polynucleotides were recognized: (a) a strong ionic type prominent in complexes of the antibiotics with the homopolynucleotides poly U, poly C, and poly A ; this binding was reversible to extensive dialysis at high ionic strength; (b) an effectively irreversible type found in all DNA-antibiotic complexes studied ; this type was partially reversed by dialysis at high ionic strength, and was characterized by the appearance of shoulders in the visible spectrum of the antibiotics; (c) an interaction of antibiotic molecules with one another to cause their aggregation on the surface of DNA. Hedamycin, which is more potent than rubiflavin as a bactericidal agent, causes a much greater change in the properties of DNA per molecule bound than does rubiflavin, although the maximum number of hedamycin molecules that can bind per D N A nucleotide is only about half that of rubiflavin. The strong affinity of these antibiotics for DNA suggests an antibacterial mechanism in which rubiflavin and hedamycin become intracellularly bound to DNA and prevent DNA synthesis by inhibiting strand separation.

physical-chemical methods show that rubiflavin binds strongly to DNA in vitro. Rubiflavin lowers the buoyant density of DNA in CsCl gradients roughly in proportion to the molar ratio of rubiflavin to DNA nucleotide present in the complex. We therefore suggested that the antibacterial properties of rubiflavin appear to be a consequence of its tendency to bind to DNA. Hedamycin, C41Hj2N20 11, isolated from Streptomyces griseoruber (Schmitz et al., 1967), has been reported to be similar to, but more potent than, rubiflavin in biological properties (Bradner et al., 1967; Heinemann and Howard, 1966). Its bactericidal action has been studied and appears to be entirely analogous to that of rubiflavin (H. L. White and J. R. White, to be published). In this paper complexes of the two antibiotics with D N A and other polynucleotides are compared. Materials and Methods The following were generous gifts: rubiflavin from Dr. A. Aszalos of the Squibb Institute for Medical

VOL.

8,

NO,

3,

MARCH

1969

with DNA, a fresh rubiflavin solution was prepared at Research; hedamycin from Dr. W. T. Bradner of Bristol Laboratories ; daunomycin hydrochloride from twice the desired final concentration by adding microliter amounts of 1000 pgjml of cold ethanolic stock soluDr. F. Arcamone of Farmitalia, Milan, Italy; nogalation to cold buffer. Rubiflavin and D N A solutions were mycin from the Upjohn Co. mixed, always adding rubiflavin to DNA, slowly and Calf thymus and Clostridium perfringens D N A were with stirring, in order to achieve uniform mixing. For purchased from Worthington Biochemical Corp. and the reference cuvet, equal volumes of D N A and salmon sperm D N A (highly polymerized) from Calbuffer were mixed in the same way. Usually, final conbiochem. Commercial D N A samples were dissolved a t centrations were 1000 pg/ml of DNA and 10 pg,'ml of M saline with 2-5 mg/ml in PE6l buffer or in gentle stirring a t 4" for 3 days. Solutions were then rubiflavin (or 5 pg/ml of hedamycin). The spectrum dialyzed exhaustively at 4" against PE6 buffer. Escherwas run on the Cary 14 spectrophotometer using a ichia coli and Micrococcus Iysodeikticus DNA were sensitive slide-wire (full-scale absorbance 0.1). This, at isolated from bacterial cultures using the method of maximum DNA concentration, represented the specMarmur (1961). Tritium-labeled D N A (1.1 X lo7 trum of fully bound rubiflavin. Samples containing higher proportions of rubiflavin were prepared by pCi/MJ was prepared from E . coli strain 15 T-Phediluting the first DNA-rubiflavin solution, slowly and grown in the presence of [methyl-3H]thymine. Likewise with mixing, with a fresh aqueous rubiflavin solution [14C]DNA (1.4 X 10" pCi/M,) was prepared by growing cells with [2-14C]thymine, and [ p ~ r i n e - ~ ~ C C ] (10 pg'ml). At the same time the DNA reference soluD N A (1.4 X 10" pCijM,) by growing cells in the tion was diluted with an amount of buffer equal to the volume of rubiflavin solution added to the sample. presence of [8- 4C]guanine. Immediately after isolation these radioactive samples were exhaustively dialyzed In this way a series of spectra was prepared at constant against standard saline citrate (0.15 M NaCl and 0.015 M rubiflavin concentration, but with decreasing DNA concentration. trisodium citrate, pH 7.0) and stored frozen in 1-ml batches at concentrations of from 0.5 to 1.0 mg per ml. T,, Determinations. DNA-rubiflavin and DNABefore use these D N A samples were exhaustively hedamycin melting transitions were determined in PE6 dialyzed against PE6 or other appropriate buffer and buffer or in PE6 buffer with NaCl added up to 0.10 M to give increased ionic strengths. Solutions for T,,,work assayed by ultraviolet absorbance. Poly U , poly C , and poly A were purchased from with rubiflavin and hedamycin were prepared as deMiles Chemical Co. ; poly dAT from Biopolymers, Inc. ; scribed above for spectrophotometry, using D N A yeast R N A from Pabst Laboratories; deoxynucleoside which had been dialyzed against the appropriate PE6 buffer. Melting transitions were determined with a monophosphates from Calbiochem; bovine plasma albumin from Armour Pharmaceutical Co. ; snake Gilford Model 2000 spectrophotometer. I n a typical run the four cuvets contained: (1) buffer, (2) DNA venom phosphodiesterase and ribonuclease from solution, (3) DNA-antibiotic solution, and (4) antiWorthington Biochemical Corp. [3H]Poly U (ammobiotic alone. Usually 40 pg/ml of DNA was present in nium salt, 27.33 mCi/mmol) was purchased from Miles 2 and 3. To study the influence of per cent guanine plus Laboratories, Inc. Buoyant Densities. Centrifugations of DNA-antiC) of DNA on melting transitions cytosine (% G biotic complexes were performed in a Beckman-Spinco obtained with antibiotics, experiments were done with Model E analytical ultracentrifuge, using an AN-D the following D N A samples: CI. perfringens (307, rotor and the 4O-sector cell with Kel-F centerpiece. G C), salmon sperm (427,), calf thymus (44y0), Techniques were substantially as described by Mandel E. coli (50Y0),and Micrococcus Iysodeikticus (727,) et al. (1968). Sueoka (1961). Before a run each cuvet was bubbled Absorption Spectroscopy. Spectra of antibiotics were with helium for 5 min and immediately covered so that determined using a Cary Model 14 recording spectroa water seal was made around the ground-glass stopper. photometer. Stability of antibiotics in various solvents After initial absorbances at 260 mp were recorded, the was studied by following changes in spectra with time full-scale absorbance of the instrument was set to give and by comparing biological activity against E. coli maximum sensitivity (0.3 full-scale absorbance for 40 with that of a freshly prepared solution. kg/ml of DNA), and the offset control was adjusted Spectral studies of interactions of rubiflavin or in order to zero all samples. Temperature was raised hedamycin with D N A were done in the following way a t OS"/min, while the instrument automatically re(outlined for rubiflavin) in order to minimize errors corded absorbance changes, A A , in each cuvet a t caused by the instability of rubiflavin spectra and the intervals of 5 sec/cuvet. A fifth channel recorded temstrong tendency of both of these antibiotics to form perature within the cuvet chamber. Transition midfibrous precipitates with DNA. Salmon sperm D N A points, T,, were calculated from replots of A A cs. (Calbiochem), after exhaustive dialysis against PE6 temperature. buffer, was diluted to twice the highest D N A conWhen absorbance no longer changed with tempercentration to be used. Immediately prior to mixing ature the circulating pump was shut off, and the instrument continued to record changes during slow cooling. Final absorbance of slowly cooled samples Abbreviations used: PE6 buffer, M sodium phosphate was read after about 15 hr to give per cent renaturation. l o w 4 M sodium EDTA brought to p H 6.0; poly dAT, the Optical rotatory dispersion spectra of salmon sperm double-stranded, alternating copolymer of deoxyadenylate and D N A in the presence of rubiflavin were determined deoxythyrnidylate; PEI, polyethylenimine.

+

+

DNA COMPLEXES OF HEDAMYCIN

1031

AND RUBIFLAVIN

BIOCHEMISTRY

1032

with a Cary Model 60 spectropolarimeter. Spectra were scanned from 550 to 200 mp with full scale = 20 mdeg, scan speed = 1 A/sec, and time constant = 10. All curves were referred to a buffer base line. Viscosity. Viscosities of D N A and DNA-antibiotic complexes were determined at 36.20' with a ZimmCrothers low shear viscometer (Beckman Instruments, Inc.), adjusted to give a shear stress of 1.33 X dyn/cm2 as calculated by the equations of Zimm (1962). Flow time for solvent alone was 182.3 sec for one turn of the rotor. At least three measurements at each concentration were averaged for specific viscosity calculations. Enzymatic Digestion of DNA. Hydrolysis of radioactive D N A was accomplished in 0.10 M Tris-C1 buffer (pH 8.8) a t 36-37' using 50 pg/ml of snake venom phosphodiesterase (Williams, 1961). Samples were removed at intervals with capillary pipets and spotted at the origin of PEI-cellulose thin-layer plates. Hydrolysis of native D N A was complete in 6 hr. Digestion products were located by autoradiography, after which portions of thin-layer plates were scraped off and counted. Details of chromatography and scintillation counting have been described (White and White, 1968). Rates of digestion of DNA-antibiotic complexes by snake venom phosphodiesterase were studied in the above Tris buffer using native or denatured [3H]DNA (40 pg/ml) at molar ratios of 0.05 and 0.10 for rubiflavin and 0.05 for hedamycin. During incubation at 36-37', 20-111 samples were spread on Whatman No. 1 filter paper disks and assayed for acid-precipitable counts per minute using a batch process (Bollum, 1966). Soluent Extraction. DNA-rubiflavin solution ( 5 ml) and organic solvent ( 5 ml) were shaken vigorously on a Burrell wrist-action shaker for 30 min and then centrifuged for 5 min at 2500 rpm. Ionic strength of the aqueous buffer was increased by adding either NaC1, CsC1, or NaC104. Exhaustice Dialysis. Since the binding of rubiflavin and hedamycin t o D N A was very strong and not of a n equilibrium type, a n exhaustive dialysis method was devised for studying the nature of this binding. D N A for these experiments was first dialyzed exhaustively against PE6 buffer at 4'. This buffer was used in all these experiments except where it was desired to increase ionic strength by adding NaCl to PE6 buffer. A series of DNA-rubiflavin or DNA-hedamycin solutions was prepared a t selected molar ratios of antibiotic to nucleotide ( r values). Usually D N A concentration was kept constant a t 4 X 10-5 M and antibiotic concentration was varied. Similar results were obtained when D N A was varied and antibiotic concentration was held constant. Solutions were allowed to remain a t 4" overnight, protected from light; 1 ml of each was then placed in a washed dialysis bag (Union Carbide 8/100) in a scintillation-type vial and dialyzed against 20 ml of buffer. A small stirring flea was placed in the bottom of each vial, and as many as 20 such vials were placed on one magnetic stirrer. These were covered with aluminum foil and kept in a cold room at about 4'. The outside buffer was changed after 1, 3, and 5 days.

WHITE AND WHITE

I n experiments with 1 M NaCl present in the buffer, solutions were first prepared in PE6, then dialyzed against PE6 containing 1 M NaCl for 6 days with three changes, and finally dialyzed against PE6 alone overnight. On day 7 the contents of each dialysis bag were assayed for antibiotic in the visible region using the Cary 14 spectrophotometer with sensitive slide-wire. D N A controls were used for spectral D N A assay and also as reference when DNA-antibiotic solutions were assayed in the visible region. E . coli, salmon sperm, M . lysodeikticus, and Cl. perfringens DNAs were studied in the above experiments. E . coli D N A was labeled either with [3H]thymine or [l4C]thymine. From spectral assays (and radioactivity measurements when E. coli D N A was used), final r values, rf, were calculated. These were plotted against initial r values, ri, to give a n indication of the strength and extent of strong binding. The homopolynucleotides poly U, poly C, and poly A were dialyzed against PE6 buffer and assayed spectrally before use in the above experiments. Molar extinctions employed, based on nucleotide phosphate, were 8000 for poly A, 8600 for poly C, and 9500 for poly U (Shugar, 1960; Michelson and Monny, 1966). Poly (rA) :poly (rU) was prepared by mixing equimolar amounts of poly A and poly U in PE6 buffer containing 0.1 M NaCl and allowing the mixture to remain at room temperature for 20 hr. At this time a spectral assay showed the expected hypochromicity ( E , = 7000) (Peacocke and Blake, 1967). Results Solubility and Stability. Rubiflavin was stable for a t least 6 weeks in absolute ethanol if kept refrigerated and protected from light. Biological activity and spectra were relatively unchanged during this time. However, as was reported to us (A. Aszalos, 1964, personal communication), rubiflavin was only slightly soluble and was very unstable in aqueous solutions: a decrease in the absorption spectrum occurred within minutes without a shift of the absorption maxima, and this decrease was accelerated by such procedures as bubbling nitrogen or air through the solution or centrifugation. Therefore aqueous solutions were prepared from a n ethanolic stock immediately before use. Sodium borate, sodium phosphate, and Tris-HCI, each at concentrations of IOv3 and lo-' M and a t p H values of 6, 8, and 9.2, were not suitable as buffers. Sodium phosphate-EDTA buffer at IOW3 M phosphate and M EDTA was studied at the same pH values and was suitable, since the rubiflavin spectrum remained stable for at least 10 min in this buffer at the three pH values. At 1 hr the absorbance was only slightly less, but at 20 hr, the visible spectrum was about 20y0 lower, although the absorption maxima were not shifted. Raising the ionic strength to 0.1 M with NaCl or raising phosphate and EDTA concentrations by a factor of 10 did not change these results significantly. Therefore, phosphate-EDTA buffer (pH 6) was chosen as a suitable solvent. The spectrum of rubiflavin was strongly dependent upon pH. A p H titra-

VOL.

8,

NO.

3,

MARCH

1969

tion indicated a pK, value of 8.1 or 8.2. Another transition near p H 10 caused a color change from amber to rosy red. Since the species present below p H 8 exhibited greater aqueous stability and biological potency, p H 6 was chosen as a n optimum p H for in citro studies. I n phosphate-EDTA buffer at p H 6.0 (PE6 buffer), stored in the dark at 4", the bactericidal activity of rubiflavin decreased gradually with time, being only 167, of its original value after 5 weeks. During this time the rubiflavin spectrum decreased, but no changes in wavelengths of maximum absorbance were observed. Hedamycin was also reported to be light-sensitive and unstable in aqueous solutions, losing 90% of its potency at room temperature in 2 hr (W. T. Bradner, 1966, personal communication). Hedamycin was only slightly soluble in ethanol and exhibited complete loss of antibacterial potency when stored (at 4" and dark) in dimethylformamide for 2 months. I n PE6 buffer, however, we found no significant change in spectrum or potency against E . coli over a period of 3 months if the solution was kept refrigerated and protected from light. Hedamycin could be dissolved directly in PE6 buffer at a maximum of 50 pg/ml, and this was used as a stock solution. Clearly hedamycin is much more stable in aqueous solutions than rubiflavin. Absorption Spectra. The aqueous absorption spectra, at p H 6, of rubiflavin and hedamycin are very similar. Both compounds have broad maxima a t 428 rnk and sharp peaks at 245 mp with shoulders near 270 mp. Hedamycin has an extra peak at 212 mp. Molar extinctions are quite different, being for rubiflavin at 10 pg/ml at 245 mp, 20,000, and at 428 mp, 3700; and for hedamycin at 245 mp, 46,000, at 428 mp, 10,000, and at 212 mp, 48,000. Molecular weights of rubiflavin and hedamycin are 412 and 748, respectively (Aszalos et al., 1964; Schmitz et al., 1966). The absorbance of neither compound follows Beer's law, but in the concentration range used in this work (0-10 pgiml), the extinctions cited above were applicable. Chromatography and Electrophoresis of Rubiflacin and Hedamycin. Chromatography of both compounds on Eastman preformed silica gel thin layers was achieved using mixtures of alcohol-organic acid-water in various ratios. In isoamyl alcohol-formic acid-water (3:2:1, v/v), rubiflavin gave a single distinct spot with RF 0.43, while hedamycin gave two distinct spots, of nearly equal ultraviolet intensity, having R F values of 0.41 and 0.32. Hedamycin may have been degraded by the acidic solvent into two components. Electrophoresis of rubiflavin and hedamycin on Whatman No. 1 paper strips in a Gelman chamber indicated that both compounds were positively charged at pH 6 and 7.4. Reducibility. Both rubiflavin and hedamycin were reduced irreversibly by sodium borohydride, rubiflavin going from an amber to a colorless solution and hedamycin from yellow to colorless. The visible maximum of rubiflavin was shifted from 428 to 306 mp. When sodium hydrosulfite was the reducing agent, the 428-mp maximum of both compounds was red shifted to near 500 mp in the visible region. Hydrosulfite reduction was largely reversible, and after standing in air

Half-Wave Potentials Electrode."

TABLE I:

Compound

E112 (V)

Rubiflavin Hedamycin

-0.36 -0.42

Nogalamycin Daunomycin

-0.42 -0.60

-1.5

L'S.

Saturated Calomel

Molar Diffusion Current (PA) 1.0 x 1036 x 103 c 1.1 x 104 3 x IO3 (uncertain) 7.9 x 103 5.0

Tris buffer, 0.10 M, p H 7.4. Solutions contained ethanol. Solutions contained 5% dimethylformamide. a

5%

for a few minutes, the solutions turned from red back to their original colors. Both compounds were reducible at the dropping mercury electrode. Half-wave potentials measured with a saturated calomel reference electrode at p H 7.4 are given in Table 1. Hedamycin was distinguished from rubiflavin by a much higher diffusion current and two distinct waves. I n order to achieve practical concentrations for polarography, rubiflavin was diluted from an absolute ethanol solution into Tris-HC1 buffer, and hedamycin from a freshly prepared concentrated dimethylformamide solution. The anthraquinone antibiotics, daunomycin and nogalamycin, were included for comparison. The latter, being very insoluble, was of uncertain concentration. Its similarity to hedamycin in polarographic behavior is of interest. Electron spin resonance spectrometry has shown that rubiflavin causes a free-radical signal to appear in cultures of E . coli strain B (White and Dearman, 1965), indicating that rubiflavin can be intracellularly reduced. The fact that the electron paramagnetic resonance signal lacks hyperfine structure suggests that rubiflavin free radicals may be bound to large molecules. M. T. Huang (unpublished data) has observed a similar signal lacking hyperfine structure from hedamycin in cultures of E. coli.

In Vitro Interaction with DNA Equilibrium Gradient Centrifugation. When solutions of ["HIDNA from E. coli were mixed with rubiflavin to give various r values and these solutions were centrifuged separately in a CsCl gradient in the Spinco Model L preparative ultracentrifuge, decreases in buoyant density were readily observed (White and White, 1967) for r values of 0.1 or more. In the Beckman-Spinco Model E analytical ultracentrifuge decreases in buoyant density were observed in CsCl gradients for complexes of both rubiflavin and hedamycin with native or denatured DNA. It was found that if DNA was placed in the same gradient with DNA-antibiotic complex, two distinct bands were obtained, corresponding to D N A alone and DNA-antibiotic complex. For salmon sperm DNA-

DNA

COMPLEXES OF HEDAMYCIN AND

1033

RUBIFLAVIN

BIOCHEMISTRY

rubiflavin at r = 0.10, the change in buoyant density due to complex formation was Ap = -0.014 g/cm3; for DNA-hedamycin a t r = 0.05, Ap = -0.077 g/cm3. The remarkable stability of bound complexes in CsCl gradients was demonstrated as follows. A series of E. coli DNA-rubiflavin solutions was prepared having r values of 0.10, 0.20, 0.30, and 0.40. Samples of these solutions, containing 1 pg of D N A each were mixed with CsCl along with 1 pg of E . coli marker DNA. Following centrifugation of such a mixture, distinct bands were observed for each of the complexes as can be seen in Figure 1. The position of the lowest band does not differ from that of uncomplexed DNA banded separately, indicating that D N A in this band has not become complexed with a detectable amount of antibiotic. Equivalent amounts of D N A were originally present in each complex. Thus it appears that a portion of each complex remains undissociated. When any of these DNA-antibiotic complexes was centrifuged by itself, one band at the position of lowered density appeared, and there was no indication that any unbound D N A was present. When the experiment of Figure 1 was carried out with freshly prepared samples, that is, with complexes prepared within 0.5 hr of addition to

I

MENISCUS

1: Microdensitometer tracing resulting from a CsCl gradient containing salmon sperm DNA and four DNArubiflavin complexes. Complexes having the values of ri indicated were prepared separately and held at 4" for 3 days before mixing with DNA in CsCl solution. Centrifugation in the Spinco Model E analytical ultracentrifuge was for 16 hr at 44,770 rpm. The experiment, which illustrates the great stability of the complexes, is described in the text. The upper tracing shows that somewhat better definition of the bands is obtained when DNA and only one complex are present in the gradient. FIGURE

1034

t BOTTOM

WHITE AND WHITE

CsC1, the equilibrium position of the native marker was raised, and all samples banded diffusely in the density region between the expected positions of free D N A and of the most highly complexed DNA. Thus in this case the complexes were rather unstable, and antibiotic molecules in some cases became free of DNA and later bound to another DNA molecule. Complexes of greater stability were formed when a solution of the complex was incubated for 2 hr at 40" prior to addition of CsC1. Observations similar to those described above for DNA-rubiflavin complexes were also made after centrifugation of DNA-hedamycin complexes. Solcent Extraction. The tenacious nature of the binding of rubiflavin and hedamycin to DNA suggested that the binding might be covalent. This is apparently not the case, as shown for rubiflavin. The antibiotic could be extracted from an aqueous layer by organic solvents such as benzene, chloroform, or chloroform-isoamyl alcohol mixtures. This extraction did not take place with DNA-rubiflavin complexes at r = 0.50 at either pH 6.0 or 8.0. However, when a monovalent salt at concentration of 1 M was present in the aqueous layer, the rubiflavin was extractable. Sodium perchlorate, CsC1, or NaCl served equally well. Precipitation of some of the D N A at the interface was apparent at these high salt concentrations. Absorption Spectroscopy. In the presence of DNA, the visible absorption spectra of rubiflavin and hedamycin were depressed, red shifted, and characterized by shoulders. Spectra of DNA-bound rubiflavin or hedamycin were dependent upon the manner of solution preparation, that is, interactions were not reversible, and therefore equilibrium binding theory could not be applied. Moreover, no isosbestic points were found. For spectral studies solutions were prepared in PE6 buffer as described in Methods in order to prevent precipitation of complex. It was found that when the molar ratio, r, of rubiflavin to DNA nucleotide was less than unity, the spectrum remained stable for hours or days. Figure 2 shows that, for salmon sperm DNA-rubiflavin, at the lowest r value, the 428-mp peak was shifted to 440 mp. An identical shifted spectrum was obtained for r < 0.01. As fresh rubiflavin was added in order to decrease D N A concentration while keeping rubiflavin a t constant concentration, the height of the spectrum increased until r = 0.04. Then further dilutions depressed but did not shift the spectrum until r approached 1.0, when it began to shift to the blue. At r = 1.0 the peak was at 433 mp, and the shoulders were lost. At larger r values the peak returned to 428 mp but remained depressed. If the solutions having r > 1 were allowed to sit for a few hours, their spectra became more intense without a shift in the maximum. Spectra from a similar experiment for DNA-hedamycin solutions are shown in Figure 3. As a solution having an initial r value of 0.006 was raised to 0.20, the spectrum remained the same, with maximum at 440 mp and shoulders similar to, but somewhat stronger than, those of bound rubiflavin. The spectrum began to show a blue shift at r = 0.40, which continued until r = 0.75, when the maximum was at 428 mp and

V O L . 8, N O . 3, M A R C H 1 9 6 9

i

r-\

HD

0.07

2.00

\

p H : 6.0

350

400

450

500

WAVE LENGTH (mp)

I

I

I

I

I

370

400

430

460

490

MILLIMICRONS

FIGURE 2 : Absorption

spectra of DNA-rubiflavin complexes. The broken line labeled rubiflavin is the spectrum of free rubiflavin. Other curves, denoted by the molar ratio, r , of rubiflavin to DNA nucleotide, are the spectra of solutions prepared as described in Materials and Methods.

FIGURE

3: Absorption spectra of DNA-hedamycin complexes. The curve labeled hedarnycin is the spectrum of free hedarnycin. Other curves, denoted by the molar ratio, r , of hedamycin to DNA nucleotide, are the spectra of solutions prepared as described in Materials and Methods.

shoulders had disappeared. A further increase in r did not change the spectrum until r = 2.0, when absorption a t 428 mp increased. D N A from E. coli, M . lysodeikricus, and Cl. perfringens gave results similar to salmon sperm DNA with both rubiflavin and hedamycin. The above spectral changes were strongly inhibited by 0.01 M magnesium ion in all solutions. Although spectra still showed evidence of an interaction, the peak did not shift to 440 mp, and no shoulders were generated. The fact that PE6 buffer tends to stabilize the rubiflavin spectrum suggests that magnesium (and other divalent cations) may interact with the antibiotic directly. However, magnesium bound to DNA phosphates might also modify the nature of rubiflavin binding to DNA. Rubiflavin which had been refrigerated in PE6 buffer for 5 weeks and which retained only about 16% of its biological activity was still capable of binding to D N A in cifro, giving a spectrum with a maximum a t 440 mp and shoulders typical of bound rubiflavin, although of higher absorbance than that obtained with fresh rubiflavin. The spectrum of DNA-bound rubiflavin was similar at p H 6.0, 7.4, and 9.1, although spectra of fresh rubiflavin solutions varied with pH. At p H 9.1 the complex tended to settle out of solution

in a yellow gelatinous form which, when shaken, gave a reproducible bound-rubiflavin spectrum. Heat-denatured DNA at p H 6.0 bound both rubiflavin and hedarnycin, and the spectra of the bound antibiotics were similar to those found with native DNA. However at p H 12.5, where DNA is denatured and rubiflavin is in its red alkaline form, no binding was observed. Reduction of rubiflavin by borohydride in the presence of D N A tended to reduce binding. The four 5’-deoxyribonucleotides were used in binding studies with rubiflavin at initial r values of 0.01. None gave spectral shifts or generation of shoulders. However some depression of absorbance was observed for 5’-dAMP and 5’-dTMP. No interaction at pH 6.0 between rubiflavin and E. coli ribosomes was detected. Likewise no binding with bovine plasma albumin was observed. With yeast R N A at r = 0.07, the spectrum of the rubiflavin complex was lower than that of free rubiflavin. The maximum was red shifted to 437 mp and exhibited shoulders similar to the spectrum of DNArubiflavin. Interactions with the homopolymers, poly A, poly C, poly U, and poly (rA) :poly (rU), also gave depressions and shifts to 437 mp, but no definite shoulders were observed. On the other hand, the alternating

D N A

1035

COMPLEXES OF HEDAMYCIN A N D RUBIFLAVIN

BIOCHEMISTRY

copolymer, poly d(AT), gave a bound rubiflavin spectrum similar to that of DNA-rubiflavin. I n organic solvents the visible absorption maxima of both rubiflavin and hedamycin were red shifted, the largest shift being obtained with benzene. These spectra are shown for benzene and ethanol in Figure 4 in the case of rubiflavin. Besides inducing red shifts, the more hydrophobic solvents caused the appearance of several shoulders, which were similar for both antibiotics. Also shown in the figure for comparison is a n aqueous spectrum of rubiflavin in the presence of excess salmon sperm DNA. The DNA-bound antibiotic spectra are similar to those of the free compounds in a hydrophobic environment. Melting Transition of DNA. Rubiflavin and hedamycin both caused a large increase in T , of DNA, while absorbances of the antibiotics alone are not significantly changed on heating. The T, of DNArubiflavin complexes was studied as a function of r value, ionic strength, and per cent guanosine plus cytosine of DNA. For DNA-antibiotic complexes the T , was considered to be the temperature at which the absorbance was the same as that of the control D N A sample at its midpoint. In the presence of rubiflavin the transition occurred over a broader temperature range, and achievement of full hyperchromicity was gradual. On cooling, DNA-rubiflavin samples tended to renature much more readily than controls. The effects on

0.04-

0.03- RUBIFLAVIN IOpcp/ml

0.02-

0.0llI

I

I

1

400

1

1

I

I

450

I

'

'

\ I

'\

500

MILLIMICRONS

1036

Absorption spectra of rubiflavin in aqueous and organic solvents. FIGURE 4:

WHITE A N D WHITE

I

-

1

- 00 z

-60

50 a

3

5 -40

5

a

s - 20 I

I

0.02

I

0.04

0.06 I

0.08

I

1

QlO

FIGURE 5 : Increase in Tm of DNA and per cent renaturation after slow cooling as a function of molar ratio of rubiflavin to DNA nucleotide. The melting transitions of salmon sperm DNA-rubiflavin complexes were studied in PE6 buffer as described in Materials and Methods. The T , of uncomplexed DNA was 50.3".

T , and per cent renaturation are illustrated as a function of r in Figure 5 for salmon sperm DNA. There was no clear dependence upon per cent guanosine plus cytosine. Variation of the ionic strength from 0.001 to 0.01 M did not significantly affect the increase in T , obtained at a particular r value. At an ionic strength of 0.10 M, the increase in T , was reduced by at least half, but in this case transitions were occurring in a higher temperature range, and the effect of temperature on the nature of the complex is uncertain. At low ionic strength (0.001 M) the effect of rubiflavin on renaturation upon slow cooling was much more pronounced, since D N A alone lost only 4% of its hyperchromicity, compared with 44% for rubiflavin at r = 0.10. At r values higher than 0.2, melting curves were multiphasic, but no plateau could be reached at the highest temperature readily attainable with a water bath (94"). With hedamycin the increase in T , of D N A was much more pronounced t h m with rubiflavin. At r = 0.025 with salmon sperm D N A at an ionic strength of 0.003 M, the increase in Tm was at least 20") but a plateau could not be reached. Since the antibiotic molecules might have become dislodged from their D N A binding sites during the heating process, visible spectra were recorded before, during, and after T , determinations. For salmon sperm DNA-rubiflavin at r = 0.17 and ionic strength 0.003 M, the spectrum obtained a t the plateau of the DNArubiflavin melting transition was similar to that of free rubiflavin, suggesting that at that temperature the DNA-rubiflavin complex was largely dissociated. On slow cooling the DNA-rubiflavin spectrum shifted about halfway back. The manner in which degradation of the rubiflavin molecule may have contributed to these observations is not known. In the case of DNA-hedamycin, at r = 0.025 and an ionic strength of 0.003 M, a plateau had not been reached at 92" where a visible absorption spectrum was

VOL.

8,

NO.

3,

MARCH

1969

6: Determination of apparent intrinsic viscosities of DNA-rubiflavin and DNA-hedamycin complexes. Molar ratio of antibiotic to salmon sperm DNA nucleotide was kept constant by dilution of more concentrated solutions of DNA-antibiotic complex with PE6 buffer. C1 = concentration of complex in micrograms per milliliter. FIGURE

I

I

IO

I

20

I

I

40

30

1

50

I

60

CI

taken. The DNA-hedamycin spectrum was blueshifted toward that of free hedamycin on heating, but it did not lose its fine structure, and on cooling it was shifted back to its original position. Optical Rotatory Dispersion of DNA-Rubiyacin (White, 1967). Rubiflavin alone exhibited multiple Cotton effects in the ultraviolet region and a broad negative Cotton effect centering near 440 mp. At a higher concentration (50 pg/ml) a positive Cotton effect a t 245 mp became more prominent than a t lower concentrations. I n the presence of DNA, optical rotatory dispersion spectra were not additive, and no extrinsic Cotton effect appeared in the visible region. The optical rotatory dispersion spectrum of D N A is quenched by bound rubiflavin (r = 0.25) at wavelengths lower than 270 mp. This effect is most striking near 200 mp where native D N A exhibits a sharp peak. Viscometry. Using the low-shear viscometer of Zimm and Crothers (1962) viscosities of DNA-rubiflavin and DNA-hedamycin complexes were studied at r values where the antibiotics could be assumed to be completely bound. By diluting with buffer, D N A concentration could be decreased while r remained constant. Plots of the reduced viscosity, vSp/c,cs. concentration of DNA showed that, although the specific viscosity of D N A was increased by both antibiotics, the extrapolated “apparent intrinsic viscosities” were not significantly different at different r values. When calculations were made with respect to weight concentration

of complex rather than of DNA, curves for three r values tended to be superimposed and to extrapolate to the same point (Figure 6). When DNA-rubiflavin complexes were diluted with rubiflavin, keeping the rubiflavin concentration constant, the extrapolated values for intrinsic viscosity were much greater, but in this case the composition of the complexes changed (that is, r increased) as the D N A concentration was lowered. Therefore n o attempt was made to interpret such results quantitatively. Enzymatic Digestion Rates and Products. Native DNA, when bound to either rubiflavin or hedamycin, was strongly resistant to digestion by snake venom phosphodiesterase. However, if D N A was denatured prior to antibiotic treatment, the complexes were much less resistant to digestion, as shown in Figure 7. Although the denatured D N A control was more readily hydrolyzed than the native DNA, this difference was even more pronounced in the presence of the antibiotics. In both instances DNA-rubiflavin is less resistant to digestion than DNA-hedamycin, even though hedamycin is present at only half the molar concentration of rubiflavin. Digestion products of native DNA-antibiotic complexes, in which D N A was labeled with both [3H]thymine and [14C]purine, were separated by thin-layer chromatography on PEIcellulose. Autoradiography of the chromatograms revealed that products of low mobility, presumed to be oligonucleotides, were formed during the enzymatic

1037

D N A COMPLEXES OF HEDAMYCIN AND RUBIFLAVIN

BIOCHEMISTRY

digestion in the presence, but not in the absence, of the antibiotics. Both rubiflavin and the hedamycin remained at the origins of these chromatograms. Dialysis. Results of exhaustive dialysis experiments with DNA-rubiflavin and DNA-hedamycin are presented in Figures 8 and 9. With native and denatured D N A all curves of rf (final r) us. ri (initial r) were composed of two linear regions, with a steep initial slope and a more gentle slope at higher rl values. The rf values at the intersection of the two linear regions probably represent irreversibly bound antibiotic. On a molar basis rubiflavin was bound more extensively than hedamycin by a factor of about 2, and this binding was not greatly different with native and denatured DNA. Neither was the binding significantly dependent upon yo G C. Dialysis against 1 M NaCl decreased the binding by a factor of about one-half for both antibiotics. Visible spectra after dialysis resembled those shown in Figures 2 and 3 for the completely bound antibiotics, with maxima at 440 mp and distinct shoulders. Results with the homopolymers, poly C, poly A, and poly U, showed less binding of an irreversible type. With hedamycin, none at all was found with poly A or poly C. The copolymer, poly (rA) :poly (rU), pre-

0 MICROCOCCUS DNA

CL. PERFRINGENS DNA DENATURED €.COLI DNA POLY u

e

A POLY C

0.6 u-

+

-Native DNA Denatured DNA

---

30

60

90

MINUTES INCUBATION

I20

FIGURE I: Rates

of digestion of E. coli DNA-rubiflavin and DNA-hedamycin complexes. Native or heat-denatured [3H]DNA was cornplexed with rubiflavin or hedamycin to give molar ratios of antibiotic to nucleotide as indicated below. Digestion by snake venom phosphodiesterase was followed by assaying acid-precipitable counts per minute in 25-pl samples. Controls contained 20,000 cpm/25-p1 sample at time zero. ldentification of curves: 0 , control DNA; A, DNA rubiflavin ( I = 0.10); 0 , DNA hedamycin ( r = 0.05); solid lines, native DNA; dashed lines, denatured

+

1038

Iso

DNA.

WHITE A N D WHITE

+

8: Exhaustive dialysis of DNA and other polynucleotides in the presence of rubiflavin. ri = molar ratio of rubiflavin to nucleotide prior to dialysis against PE6 buffer; rf = molar ratio after exhaustive dialysis as described in Materials and Methods. E. coli [14C]DNA was assayed in each solution before and after dialysis by scintillation counting. Other polymers were assayed by ultraviolet absorbance in controls which contained no rubiflavin. The Em of rubiflavin bound to DNA is 2340 at 440 mp. This value was used as an approximation for rubiflavin bound to other polynucleotides. FIGURE

pared from poly A and poly U, was studied only with rubiflavin and gave results similar to those for poly A alone. On the other hand, the alternating copolymer, poly d(AT) gave an rf us. ri curve quite similar to that obtained with DNA. With all these synthetic polymers except poly d(AT), spectra after dialysis were different in appearance from those found with DNA, in that they were broad smooth curves with no apparent shoulders. Maxima were at 437-440 mp. Because interactions between DNA and hedamycin or rubiflavin seemed somewhat similar to'those found by others with nogalamycin and daunomycin (Kersten et al., 1966; Calendi et al., 1965), these two anthraquinone antibiotics were compared in exhaustive dialysis experiments with labeled E . coli DNA. Figure 10 shows that they give rf us. ri curves which are similar in character to those obtained with rubiflavin and hedamycin. The values of the initial slope, and the values at which the slope changed, rfx, are listed in Table I1 for all of the above dialysis experiments. These values give a rough indication of strength of binding and extent of binding, respectively. Discussion Rubiflavin and hedamycin behave as distinctly different compounds as shown by their solubility characteristics, aqueous stability, chromatographic behavior, molar extinctions, and half-wave potentials. However, they are similar in absorption spectra in

VOL.

8,

NO.

3,

MARCH

1969

0.25

9: Exhaustive dialysis of DNA and other polynucleotides in the presence of hedamycin. Buffer was PE6, except where 1.0 M NaCl was added to PE6 as indicated. ri = initial molar ratio of hedamycin to nucleotide; rf = final molar ratio of hedamycin to nucleotide. E. coli [l4C]DNA was assayed in each solution before and after dialysis by scintillation counting. Other polymers were assayed by ultraviolet absorbance in controls containing no hedamycin. The concentrations of hedamycin were determined by visible spectra. The E , of hedamycin bound to DNA is 6880 at 440 mk. This value was used as an approximation for hedamycin bound to other polynucleotides.

FIGURE

0.20

0.15 ’ L

0.10 A

#--i=-i

NATIVE DNA

0.05

aqueous and nonaqueous solvents, in their response t o reducing agents, in ionic charge, and in antibacterial spectra. The structural differences between them seem to confer upon hedamycin greater aqueous stability and antibacterial potency. Both hedamycin and rubiflavin show a strong tendency to bind to DNA. This binding probably comprises a t least three general types: (a) an ionic interaction, probably with phosphate groups, which is diminished at high ionic strength; (b) a nonionic, essentially irreversible binding that is revealed by exhaustive dialysis experiments at high ionic strength and by cesium chloride gradient studies; and (c) a n aggregation of antibiotic molecules through interactions among themselves, with the nucleic acid providing a surface for the interaction. The irreversible type of binding is typical of D N A irrespective of per cent guanosine plus cytosine. It may be slightly greater with denatured DNA. The spectra of DNA-antibiotic solutions before or after exhaustive dialysis are always characterized by distinct shoulders, similar to those obtained by dissolving the antibiotics alone in a hydrophobic solvent (Figure 4). The electronic transition which gives rise to the 428-mp band in the hedamycin or rubiflavin spectrum is assumed to be a n n + T * transition, since the 428-mp band is blue shifted in going from benzene to hydrogen-bonding solvents (Sidman, 1958). The energy change involved in the wavelength shift from 442 mp in benzene to 428 mp in aqueous buffer is 2.24 kcal/mole. The shoulders in benzene are due to vibrational contributions and d o not appear in aqueous antibiotic solutions because hydrogen-bonding interactions with water tend to eliminate them. The fact that these shoulders are present in aqueous DNA-antibiotic solutions and also in RNA-antibiotic solutions is therefore evidence that the chromophores in the bound antibiotic molecules are in a nonaqueous environment. Since 1 M sodium chloride causes the extent of DNA-antibiotic binding

0.25

0.50

ri t o be reduced to about half, it appears that both nonionic and ionic interactions contribute to the “irreversible” binding at low ionic strength. A hint that nucleotide sequence may play an important part in the strong, nonionic type of binding comes from the studies with poly A, poly U, and poly C. The complexes of rubiflavin and hedamycin with these homopolymers, while showing red shifts and absorbance depression, give smooth, broad spectra without

I

I 07

-

06 -

-

0.5

X

I

05

15

IO

20

r, 10: Exhaustive dialysis of DNA-nogalamycin and DNA-daunomycin complexes. E. coli [ I 4C]DNA was assayed in each solution by scintillation counting. Free and bound antibiotics were assayed by visible absorbance. DNA-hedamycin is included for comparison.

FIGURE

1039

D N A COMPLEXES OF HEDAMYCIN A N D RUBIFLAVIN

BIOCHEMISTRY

to each nucleotide and one hedamycin to every pair of nucleotides, before aggregation begins to take place. Exhaustive dialysis experiments (Table 11) indicated that the “irreversible” component of this binding in PE6 buffer is about half the above values for both antibiotics. It is worth noting that hedamycin, although more potent than rubiflavin in biological action, binds to a lesser extent than rubiflavin. The data for rf/ri in Table I1 indicate that while binding by both antibiotics to D N A is effectively irreversible, the maximum r value achieved with hedamycin is less than half that for rubiflavin. However, one bound hedarnycin molecule decreases the buoyant density of DNA in cesium chloride considerably more than one bound rubiflavin molecule. Hedamycin also has a greater influence than rubiflavin on the D N A melting transition. Both antibiotics cause a strong stabilization of the double helix, hedamycin increasing T,, by about six times as much as rubiflavin on a molar basis. Spectra determined at high temperature suggest that heat dissociates the DNArubiflavin complex and that the D N A strands d o not completely separate until this dissociation occurs. The DNA-hedamycin complex appears to be the more heat stable, and perhaps this is why a plateau in the melting transition is not reached for DNA-hedamycin. The hyperchromicities obtained with both rubiflavin and hedamycin indicate that hydrogen bonds are broken, but strand separation is inhibited, resulting in a gradually decreasing slope in melting transitions as the temperature is raised (Ginoza and Zimm, 1961; Crothers and Zimm, 1964). The rate and extent of digestion of the DNA-antibiotic complexes by snake venom phosphodiesterase further illustrate the ability of rubiflavin and hedamycin to stabilize the double helix. Digestion by this exonuclease proceeds from the 3’ end of DNA and produces 5’-deoxynucleotides (Williams, 1961). The hy-

the obvious shoulders which characterize the nonionic binding. Also, homopolymer-antibiotic binding is greatly diminished during exhaustive dialysis. Binding of rubiflavin with poly (rA): poly (rU), which results in spectra similar t o those for poly A-rubiflavin, disappeared after exhaustive dialysis against 0.1 M saline. The homopolymers, then, appear to interact with rubiflavin and hedamycin mainly by ionic mechanisms, which are reversible during exhaustive dialysis at sufficiently high ionic strength. Thus the nucleotide sequence in DNA, RNA, and poly d(AT), all of which give spectra with distinct shoulders, may play a role in the nonionic binding. The series of visible spectra determined a t constant rubiflavin or hedamycin concentration and decreasing D N A concentration apparently reflect all three kinds of binding mentioned above. At low r values spectra are red shifted and characterized by shoulders, suggesting the nonionic type of binding. I n the case of DNA-rubiflavin (Figure 2), an initial increase in absorbance at 440 ml.r occurs with increasing r up to r = 0.04, when one rubiflavin was present for every 25 nucleotides. At this point spectra begin to decrease as though stacking of rubiflavin along the helix might be causing interactions between adjacent rubiflavin molecules (Michaelis, 1947; Bradley and Wolf, 1959). This effect is not obvious in the case of the hedamycin molecule (Figure 3), which because of its larger size may not have the same tendency to stack upon itself. As r increases (to about 1 for rubiflavin or 0.5 for hedamycin complexes), spectra become smoother (disappearance of shoulders) and blue shifted toward the wavelengths of the antibiotics alone. At high r values the antibiotic molecules probably aggregate on the surface of the D N A molecules. Since spectral shifts in the above experiments occur at r values near 1.0 for DNA-rubiflavin and near 0.5 for DNA-hedarnycin, it may be concluded that, under the experimental conditions, one rubiflavin can bind

TABLE 11:

Parameters from Exhaustive Dialysis Experiments. Rubiflavin Polymer

DNA, E. coli, native DNA, E. coli, native (1 M NaCl) DNA, E . coli, denatured DNA, M . fysodeikticus, native DNA, Cf.perfringens, native Poly u Poly A Poly c Poly d(AT) Poly (rA):poly (rU) Poly (rA) :poly (rU) (0.1 M NaCI)

Hedamycin

n/ri

Tix

rdri

r/fx

1.o

0.53 0.23 0.57 0.57 0.42 See Figure 8 0.13 0.13 0.35 0.30

0.80

0.20

0.76 0.88

0.10 0.25

0.04

See Figure 9

0.75 0.80 0.84

0.92 0.09 0.25 0.25 0.77 0.24 0

0 0

0

Daunomycin

1040

DNA, E. coli, native

WHITE AND WHITE

0.61

0 0

Nogalamycin 0.16

1.o

0.27

VOL.

8,

NO.

3,

MARCH

1969 D N A synthesis at low concentrations, are more likely situated in the large groove of DNA, which may be the location of D N A polymerase during active D N A synthesis (Reich, 1964). The strong in citro interactions between D N A and hedamycin and rubiflavin appear to have their counterpart in cico. Bacterial D N A synthesis is specifically inhibited at low but lethal concentrations of the antibiotics, and DNA-antibiotic complexes can be isolated from E . coli cells which have been killed by the antibiotic (White, 1967; White and White, 1967, and in preparation). We suggest the following mechanism of action : the antibiotics bind strongly to intracellular D N A by forces having both ionic and nonionic components ; this binding prevents DNA strand separation, which is a requirement for the semiconservative mechanism of DNA synthesis, and the cell, being unable to replicate its DNA, is no longer viable. The greater effectiveness of the hedamycin molecule, both in cirro and in bactericidal action, is probably related to its greater size, which would enable it to extend itself over a proportionately larger section of the helix, perhaps allowing it to interact with more sites on both D N A strands.

drolysis of native D N A was strongly inhibited when complexed with either rubiflavin or hedamycin. However, when D N A was denatured before antibiotic addition, much less inhibition was observed. This implies that the antibiotics had blocked base-pair separation required for hydrolysis of native D N A by this enzyme. The initial small amount of hydrolysis which occurred with native DNA-antibiotic complexes might then have resulted from digestion of each D N A molecule from the 3’ end u p to the point where a rubiflavin or hedamycin molecule was found. These compounds may similarly block strand separation required during D N A synthesis. Because the aqueous DNA-rubiflavin complexes are dissociated in high salt solutions by shaking with benzene, a covalent type of cross-linking is not likely. With regard t o the possibility of intercalation of antibiotic between bases of DNA, increases in “apparent” intrinsic viscosity were of relatively small magnitude when compared with increases observed for proflavine and acridines, which are thought to bind by intercalation (Drummond et a/., 1966; Lerman, 1961). Therefore, on the basis of the viscosity results, intercalation does not appear to be a prominent mechanism for rubiflavin and hedamycin. The increase in molecular weight of D N A caused by bound antibiotic may be sufficient to cause the small increases in “apparent” intrinsic viscosity observed with these two compounds. However, the ligand actinomycin D has very recently been shown by Muller and Crothers (1968) to increase the intrinsic viscosity of DNA, provided that the latter is of low molecular weight. Accordingly these workers have reclassified actinomycin D as a n intercalating antibiotic. Since the effect of molecular weight of D N A on the intrinsic viscosity of DNA-hedamycin and DNA-rubiflavin complexes has not been studied, it remains possible that intercalation contributes to the nonionic component of the binding. However, intercalation is not a necessary requirement for a n effective noncovalent cross-linking of DNA chains if rubiflavin and hedamycin are able to bind in one of the grooves of the double helix by strong ionic and nonionic interactions. Daunomycin and nogalamycin are anthracycline antibiotics which bind to DNA to cause T,,, increases and buoyant density decreases in CsCl gradients (Kersten er a/., 1966). At low ionic strength they exhibit a kind of irreversible binding to D N A similar to that of rubiflavin and hedamycin, as revealed in exhaustive dialysis experiments (Figure 10). Nogalamycin is very similar to hedamycin in molecular weight, polarographic behavior, and nature of binding to DNA. However, the ultraviolet and visible spectra of both nogalamycin and daunomycin (Calendi et a/., 1965) are entirely different from those of rubiflavin and hedamycin. Also they are reported to be specific inhibitors of RNA synthesis rather than of DNA synthesis (Ward er a / . , 1965; Bhuyan and Smith, 1965), and therefore might be bound along the minor groove of DNA, as is indicated in the case of actinomycin (Reich and Goldberg, 1964). Rubiflavin and hedamycin, on the other hand, because of their specific effect on

Acknowledgment The authors thank Mr. Thomas Owens Vaughan for operation of the analytical ultracentrifuge.

References Aszalos, A., Jelinek, M., and Berk, B. (1965), Antimicrobial Agents Chemotherapy 1964, 68. Bhuyan, B. K., and Smith, C. G. (1965), Proc. Narl. Acad. Sci. U. S. 54, 566. Bollum, F. J. (1966), in Procedures in Nucleic Acid Research, Cantoni, G . L., and Davids, D. R., Ed., New York, N. Y., Harper & Row, p 296. Bradley, D. F., and Wolf, M. K. (1959), Proc. Narl. Acad. Sci. U . S. 45, 944. Bradner, W. T., Heinemann, B., and Gourevitch, A. (1967), Antimicrobial Agents Chemotherapy 1966,613. Calendi, E., Di Marco, A., Reggini, M., Scarpinato, B., and Valentini, L. (1965), Biochim. Biophys. Acta 103, 25. Crothers, D. M., and Zimm, B. H. (1964), J . Mol. Biof. 9, 1. Drummond, D. S., Pritchard, N. J., SimpsonGildemeister, V. F. W., and Peacocke, A. R . (1966), Biopolymers 4, 971. Ginoza, W., and Zimm, B. H. (1961), Proc. Natl. Acad. Sci. U. S. 47, 639. Heinemann, B., and Howard, A. J. (1966), Antimicrobial Agents Chemotherapy 1966, 488. Kersten, W., Kersten, H., and Szybalski, W. (1966), Biochemistry 5, 236. Lerman, L. S. (1961), J . Mol. Biol. 3, 18. Mandel, M., Schildkraut, C. L., and Marmur, J. (1968), Methods Enzymol. 12B, 184. Marmur, J. (1961), J . Mol. Biol. 3, 208.

D N A

1041

COMPLEXES OF HEDAMYCIN AND RUBIFLAVIN

BIOCHEMISTRY

Michaelis, L. (1947), Cold Spring Harbor Symp. Quant. Biol. 12, 131. Michelson, A. M., and Monny, C. (1966), Proc. Natl. Acad. Sci. U. S. 56, 1528. Muller, W., and Crothers, D. M. (1968), J. Mol. Biol. 35, 251. Peacocke, A. R., and Blake, A. (1967), Biopolymers 5 , 383. Reich, E. (1964), Science 143, 684. Reich, E., and Goldberg, I. H. (1964), Progr. Nucleic Acid Res. 3, 183. Schmitz, H., Crook, K. E., and Bush, J. A,. (1967), Antimicrobial Agents chemotherapy 1966, 606. Shugar, D. (1960), 3, 39. Sidman, J. W. (1958), Chem. Reu. 58, 689.

Sueoka, N. (1961), J . Mol. Biol. 3, 31. Ward, D. C., Reich, E., and Goldberg, I. H. (1965), Science 149, 1259. White, H. L. (1967), Ph.D. Dissertation, University of North Carolina, Chapel Hill, N. C. White, H. L., and White, J. R. (1967), Antimicrobial Agents Chemotherapy 1966, 227. White, H. L., and White, J. R. (1968), Mol. Pharmacol. 4, 549. White, J. R., and Dearman, H. H. (1965), Proc. Natl. Acad. Sci. U. S. 54, 887. Williams, E. J. (1961), J . Biol. Chem. 236, 1130. Zimm, B. H. (1962), Proc. Natl. Acud. Sci. U. S . 48,905. Zimm, B. H., and Crothers, D. M. (1962), Proc. Natl. Acad. Sci. U. S . 48, 905.

Evidence for a Phosphoryl-Enzyme Intermediate in Alkaline Phosphatase Catalyzed Reactions* Harold Barrett,? Roy Butler,$ and Irwin B. Wilson

Kinetic evidence is presented which demonstrates the formation of a catalytic phosphoryl-enzyme intermediate during the hydrolysis of phosphate esters by alkaline phosphatase. Nine phosphate esters were hydrolyzed in the presence of 1 M Tris acting as a phosphate acceptor in competition with water and the ratios of the products were measured. Precisely 1.39 equiv of 0-phosphoryl Tris was formed for every equivalent of Pi regardless of the particular ester that was hydrolyzed. The phosphoryl-enzyme theory proposes that a phosphoryl-enzyme intermediate is formed as a step in the hydrolysis of phosphate esters followed by reaction of this intermediate with water and with other phosphate acceptors to regenerate enzyme and produce Pi and ABSTRACT:

1

he most important question concerning the mechanism of alkaline phosphatase is whether a catalytic phosphoryl-enzyme intermediate is formed during the hydrolysis of phosphate esters. Widespread interest in this question was greatly stimulated by the remarkable observations of Engstrom (1961, 1962a,b, 1964), Engstrom and Agren (1958), Schwartz and Lipmann (1963), Schwartz (1961), Pigretti and Milstein (1965), and

1042

* From the Department of Chemistry, University of Colorado, Boulder, Colorado 80302. Receiwd Ocrober IO, 1968. This work was supported by Grant NB 07156-02, National Institutes of Health, and Grant No. GB-7904, National Science Foundation. t Special Research FeIIow, National Institutes of Health 7-F3-GM-8449-02. National Science Foundation summer research participant (Grant No. GY-2443).

B A R R E T T , BUTLER, A N D W I L S O N

transphosphorylation products of other acceptors. In this theory, the leaving group of the ester is no longer present when the reaction with Tris and with water takes place and therefore cannot influence the ratio of products. Under these circumstances, the ratio of products must be a constant, independent of the actual ester which is used as a substrate. It is also true that a different ratio of products must be obtained for each ester if these reagents react with an entity which still contains the different leaving groups. Since a constant ratio of products was obtained, it may be concluded that a phosphoryl-enzyme occurs as an intermediate in the enzymic hy’drolysis of phosphate esters.

Milstein (1964), who obtained a phosphoprotein containing a phosphorylated serine side chain by incubating the enzyme (and contaminating proteins) with low concentrations of 32P-labeled Pi in acid pH. There is a tendency t o assume that this result, the formation of a phosphoprotein, proves the formation of a transient phosphoryl-enzyme intermediate during enzymic catalysis. However, reflection indicates that this observation in itself does not give us any information on this question. First, we note that the phosphoprotein is thermodynamically very stable; it is l o 5 times more stable than 0-phosphorylserine (Vladimirova et al., 1961) and 0-phosphorylethanolamine derivatives (Wilson and Dayan, 1965). Alkaline phosphatase, of course, catalyzes the synthesis as well as the hydrolysis of phosphate esters, the direction being